Search papers, labs, and topics across Lattice.
This paper introduces SCOUT, a dynamic framework for adaptive detector allocation in prompt-injection defense, which leverages historical performance data to predict the reliability and latency of various detectors for each request. By reframing the detection process from a fixed pipeline to a more flexible allocation strategy, SCOUT significantly enhances safety and efficiency, achieving a 46% reduction in attack success rate and a 40% decrease in total wall-clock time on the SCOUT-450 benchmark. Additionally, SCOUT demonstrates its effectiveness across three external benchmarks, indicating its robustness and potential for broader application in prompt-injection defenses.
A dynamic allocation strategy for prompt-injection detection can cut attack success rates by nearly half while improving response times.
Prompt-injection detectors are heterogeneous: each is strong on a different slice of attacks, and none is always reliable. Yet existing systems still treat detection as a fixed single-detector pipeline, committing every request to one detector's blind spots. We reframe defense as detector allocation: given a heterogeneous pool, decide per request which detectors to run and whether to escalate to an LLM judge. Our framework SCOUT (Scalable and Controllable Outcome-prediction for Uncertainty-aware Triage) makes this decision dynamic by predicting each detector's per-sample reliability and latency from how it behaved on similar past inputs, and exposes a single safety-utility threshold to the operator (where utility bundles benign-pass rate and wall-clock). To evaluate this setting, we build SCOUT-450, a benchmark that captures the structurally complex, agent-facing injections that older prompt-injection sets under-represent. On SCOUT-450, a safety-oriented operating point reduces attack-success rate by 46% and total wall-clock by 40% relative to an always-on GPT-4o judge, at a 5.1-point benign-utility drop. SCOUT also transfers to three external benchmarks (BIPIA, IPI, and IHEval), improving the safety-utility frontier.
